Heart Disease Prognosis Using Machine Learning Classification Techniques

Abstract

Heart Disease is one of the fatal causes of worldwide death in also third world countries like Bangladesh. Though heart disease prediction with a satisfactory accuracy level is a very demanding and challenging topic, it is achievable using advanced machine learning (ML) techniques. Manufacturing a proper ML system can not only predict cardiovascular disease with high accuracy but also can reduce human intervention, the requirement of extra medical tests. Quick prediction can deduct the death rate and severity of the disease. This paper describes our proposed methodology of predicting heart disease that targets a goal of finding important features by applying ML algorithms resulting in improving the accuracy of the prediction. Instead of collecting the dataset from the online repository, we collected data from the Sylhet region of Bangladesh by physically going to door-to-door hospitals and healthcare industries to make an appropriate questionnaire and to produce the most valuable dataset related to heart disease prediction. Our dataset consists of 564 instances and 18 attributes. We trained our model using classification algorithms like Decision Tree, Logistic Regression, K-Nearest Neighbors (KNN), Naive Bayes, Support Vector Machine (SVM), etc. Though accuracy for different algorithms varies for a different number of instances in the dataset, SVM yielded the best performance with an accuracy level of 91% for the threshold instances of the dataset in our proposed system.